Sie sind auf Seite 1von 7

Geotechnics for Sustainable Development - Geotec Hanoi 2011, Phung (edt).

Construction Publishing House ISBN 978-604-82-000-8

Design of multi-anchored walls for deep excavations

Phung Duc Long

WSP Vietnam, Hanoi, Vietnam. E-mail:

Keywords: deep excavation, multi-anchored wall, numerical analysis, Plaxis, Rankine, earth pressure

ABSTRACT: Multi-propped/anchored walls for supporting deep excavations are widely used in under-
ground construction. Design of multi-propped/anchored walls for deep excavations is a one of the most
difficult tasks for geotechnical engineers. It is a complicated soil-structure interaction problem. The pres-
ence of, among others, anchor prestress loads makes the problem more complicated. The traditional me-
thods in many cases cannot provide a satisfactory tool for design of multi-propped walls. In this paper,
results from FEM analyses of sheet pile walls for deep excavations are discused and compared with the
traditional method. The use of the finite element method, as PLAXIS, makes it possible to design the op-
timal configuration of a multi-propped wall.


Design of braced/anchored sheet piles is one of the
most complicated soil-structure problems in the Sheet piling walls are often calculated according to
geotechnical field. Simplified analytical methods traditional simplified methods using for example
cannot answer complicated design questions. Anc- the classic concept of Rankine on active and pas-
hor prestress load is one example. Sheet piling sive earth pressures. In this concept the soil behind
walls are often calculated according to the classic the wall behaves actively, and the soil at the exca-
concept of Rankine on active and passive earth vation side behaves passively. However, with anc-
pressures. However, with anchor pre-stressing hor pre-stressing loads, the classic concept of Ran-
loads, the classic concept of Rankine is no longer kine active and passive sides is not longer correct.
correct. Prestress load is used for reducing wall and In reality, the soil at the so-called active side, ac-
soil movement. However it is not easy to choose an cording to the Rankine concept, just behind a pre-
“optimal” prestress load, which can both reduce stressed anchor behaves passively.
displacements and bending moment in the wall as The first example in this paper is the excavation
well as shear stress in the dowels. Another prob- for Fredriksberg Station in the central of Copenha-
lem, the traditional simplified methods often over- gen city. The construction of a mini-metro in the
estimates bending moment in sheet pile walls. In city of Copenhagen started in November 1996. The
such cases, the design is on the safe side. However, 20.5-km system opened between 2002 and 2007.
in a certain case the bending moments are underes- The metro has 22 stations, of which 9 are
timated by the traditional methods. underground. In 2009, the metro carried 50 million
In this paper, some particular deep excavation passengers. Fredriksberg, one of the underground
projects designed by the Author using the finite stations, is located at a very dense populated area.
element method (FEM) with PLAXIS are presented. The tunnel and station is constructed by Cut &
Through these examples the Author tries to discuss Cover method. The excavation for the project is
above-mentioned aspects and to illustrate the advan- more than 200m long, 8 to 12m in depth and 25 to
tages of FEM-analysis to the traditional methods. 30 m in width.

and spacing between 2 and 4m. For the wall section
in study, three anchor levels were used at levels
+9.5, +6.0 and +2.5, i.e. at depths -1.5m, -5m and -
8.5m under the ground surface.
Both Plaxis and the Rankine method were used
for analysing. Using Plaxis, drained analysis is per-
formed. The pile wall is not watertight. The ground
water can therefore flow through the pile wall.
Two different analyses were done. In the first anal-
ysis, a prestress load of 40 kN/m was applied for
the upper anchor level and 75kN/m for the two
lower anchor levels. In the second analysis the pre-
Figure 1. Overview of the drilled-pile wall at Fredriks- stress load were 80, 150 kN/m respectively.
berg Station, Copenhagen
Table 1. Soil parameters - Fredriksberg Station,
In Fig. 1, an overview of the construction site is Copenhagen
shown. On the North side, i.e. the right side of the
Soil Fill Upper Sand Lower
figure, there is a new shopping centre, Frederiksberg clay till till clay till
Centre, which is a quite heavy five-storey building
Y-top, m +11.0 +10.0 +4.0 +1.0
located only 0.3m from the pile wall. Vertical and
horizontal displacements of the building are re- Model MC MC MC MC
stricted to 5 mm by Client. On the South side of the Mtype DR DR DR DR
project is the Old Frederiksberg station that was γ-dry, kN/m 3
16 18 18 19
built in 1864, the oldest railway station in Denmark. γ-wet, kN/m3 20 22 22 23
It had to be protected during the construction of ν' 0.3 0.33 0.3 0.33
the Metro. These two existing buildings made the E'-layer, MPa 11.1 20.8 88.8 108.7
excavation a very difficult and challenging founda- E'-increase, MPa/m 0.0 15 13.8 15
tion-engineering work, Phung (2001a). c’ - layer, kPa 1.0 1.0 1.0 1.0
The ground surface level at the studied section is c’- increase, kPa/m 0.0 0.0 0.0 0.0
+11.0 m. The soil profile includes the fill layer of 1 φ' (°) 30 33 38 35
m in thickness, the upper clay till layer 6 m thick,
the layer of sand/sand till 3 m thick and the lower A A

clay till layer with a thickness of about 7m. The clay

till is heavily over-consolidated and has a cu-value
of 150kPa, for the upper layer, and 300kPa, for the
lower layer. The limestone is situated at level –
6.0m. The modulus of compressibility of the clay till
and sand till layers increases with depth and de-
pends on the vertical effective stress. The ground
water is situated at level +2.5 m, i.e. 8.5 m below
the ground surface. In order to make it easier for
comparing the FEM analysis with the traditional
Rankine method, the c’-value is ignored for clay till. Figure 2. Drilled-pile wall at Fredriksberg Station -
Plaxis mesh
The soil properties are shown in Table 1.
A drilled-pile wall, a quite flexible wall, was The results from Plaxis analysis show clearly that
used. A pile consists of a steel tube 194-mm in di- the earth pressure at the so-called “active side” ac-
ameter and 6.3 mm thick and a steel HEB-100 core- cording to the Rankine concept just behind the pre-
beam. Every second or third tube is first drilled into stressed anchors is much larger than those in the sur-
the limestone and then filled with the HEB-beam, rounding area. Figure 3 shows that the larger the pre-
which is drilled further into the limestone. The stress loads are, the larger earth pressure can be seen.
space between the tube and the HEB-beam is filled It is obvious that there is a passive zone at the “ac-
with concrete. The soil between the piles is shot- tive” side. In this figure the effective earth pressures
creted. The pile wall is supported by injected cable obtained from Plaxis analysis is also compared with
anchors of type Supa-lina, with a declination of 30° the results from the Rankine method.

Figure 3. Effective normal stress at wall interface (effective earth pressure)

3. OPTIMAL ANCHOR PRESTRESS LOADS exchange rate. The link was built primarily to
decrease traffic in the centre of Stockholm.
Pre-stress anchor loads are often applied for control The excavation under study is about 400m long,
of the wall movements as well as displacements of 12-17m deep and 40-50m in width. The section in
the surrounding soil. It is not easy to choose the study is 17m deep. The sheet pile is of LX32 type,
“optimal” pre-stress loads. In most standards, it is driven to the bedrock. Anchors of type Dyform
only simply advised that tiebacks are prestressed to 7∅15.2 mm or 9∅15.2 mm were placed at five dif-
about 120 percent of design load and locked off be- ferent levels +13.5, +10.0, +7.0, +4.0 and +1.0, see
tween 75 and 100% of design load. The design load Figure 4.The anchors are drilled to the bedrock
is in its turn calculated from the active earth pres- with an inclination of 45°. Configuration of the
sures, according to the classic concept of Rankine sheet pile wall is shown in Fig. 5.
active and passive sides. As mentioned above with At the studied section, the ground surface level
the presence of anchor pre-stressing loads, the Ran- is +15.0m. The soil profile includes a fill layer of
kine concept is no longer correct. In reality, the soil 1.5m thick, the upper clay layer 3.5m, and the low-
at the so-called “active side”, behind a pre-stressed er clay layer 12m. The upper clay layer extents to
anchor will behave passively. +10m with constant properties. The lower clay
This also means that larger applied prestress loads layer has undrained shear strength and deformation
may cause larger earth pressure behind the wall. The modulus linearly increasing with depth. Bedrock is
advice to use a prestress load off between 75 and found at a level of –2.0m. Soil parameters are
100% of design load is perhaps a bit confused. summarised in Table 2. The ground water level is
In the second example in this paper, the sheet- +13.5m, i.e. 1.5m below the ground surface.
pile wall for the project SL-10, Södra Länken The staged construction is simulated. The sheet
(South Link) in Stockholm, Sweden, is discussed. piles are first driven to the bedrock and the first
The link is 6 km in length, of which 4.7 km is in excavation is made to a level of +13.5m, i.e. 1.5 m
tunnels, Phung (2001b). This makes it the second under the ground surface. The first anchor level is
longest urban motorway tunnel in Europe after then installed with a pre-stressing load of
Madrid M30 orbital motorway. The construction of 200kN/m. The second excavation is made to a lev-
Södra länken began in 1997, and was inaugurated el of +10.0 m, i.e. 5 m under the ground surface.
in October 2004. The total cost was about 7.9 Anchors at the second level are then installed with
billion SEK, or 1 billion USD at the 2003-2004 a pre-stressing load of 320kN/m. The process is

continued to the final excavation at a level of – Anchor 1

2.0m or 17m under the ground surface. During 600 Anchor 2

Final anchor load (kN/m)

each excavation phase the ground water is lowered Anchor 3
Anchor 4
to the excavation bottom, which is simulated by a Anchor 5
groundwater flow analysis for calculating the new
pore water pressure distribution.

Table 2. Soil parameters - South Link SL10, Stockholm

Soil Fill Upper clay Lower clay 200

Y-top, m +15.0 +13.5 +10.0

Model MC MC MC
Mtype UN UN UN 0
γ-dry, kN/m3 18.0 17.5 17.5 0 25 50 75 100 125 150
Prestress load Ratio PR (%)
γ-wet, kN/m3 21.0 17.5 17.5
ν' 0.3 0.3 0.3 Figure 6. SL-10, South Link, Stockholm - Final anchor
E'-layer, MPa 6.0 4.4 4.4 loads versus PR-ratio
E'-increase, 0.0 0.0 1.4
c’ - layer, kPa 1.0 16.0 16.0
c’- increase kPa 0.0 0.0 2.0 Maximum bending moment (kNm/m)
φ' (°) 35.0 0.0 0.0



0 25 50 75 100 125 150
Prestress load Ratio PR (%)
Figure 7. SL-10, South Link, Stockholm - Maximum
wall bending moment versus PR-ratio

Figure 4. Overview of sheet pile wall at project SL-10,
Maximum Shear Force (kNm/m)

South Link, Stockholm

+15.0 sheet pile LX32
+13.5 GWL


+ 7.0



+ 4.0 400


.2 )




.2 )

+ 1.0

.2 )



.2 )

Final excavation - 2.0


Bed rock 0 25 50 75 100 125 150
Prestress load Ratio PR (%)
Figure 5. Sheet pile wall at SL10, South Link, Stock-
holm - Configuration of the problem Figure 8. SL-10, South Link, Stockholm - Maximum
wall shear force versus PR-ratio

Figure 9. SL-10, South Link, Stockholm - Wall displacement in different studied cases

A parameter study for seven cases with different minimise the wall and soil movement, bending
pre-stress loads was performed using Plaxis. The moment, as well as shear force in the wall.
case with prestress load 200 kN/m for the first anc- With increasing PR ratio, the maximum wall
hor and 320 kN/m for other lower anchors will be bending moment Mmax and the maximum wall
used as a reference case, which is called Case 100%. shear stress Qmax decreases considerably, see Fig-
In other cases, prestress loads will be 0%, 25%, ures 7 and 8. This means that the large the prestress
50%, 75%, 125% and 150% of those in the refer- load is applied, the lower the bending moment and
ence 100% case. Let us define the prestress load ra- the shear force in the wall are. The maximum dis-
tio, PR, as the ratio in percent between the applied placement of the wall also decreases considerably
prestress load and the reference prestress load, Case with increasing prestress loads, but reaches to a
100%. This means that in Case 50% for example, a minimum value at PR= 125%, afterward it increas-
prestress load of 100 kN/m is applied for the first es. However at this point PR= 125%, a large back-
anchor and 160 kN/m for other anchors. ward displacement has happened at the wall top,
In this paper, only the behaviour of the wall at which may be dangerous for the adjacent existing
the final stage is studied. The anchor loads at the constructions, see Figure 9.
final construction stage in all cases are drawn ver- Which are the “optimal” prestress anchor loads?
sus PR-ratio in Figure 6. It can be seen that even in We can define them as the loads that cause mini-
Case 0%, i.e. no prestress load is applied for any mum wall movements and at the same time the
anchor, the final anchor loads are quite compara- lowest cost for the wall construction. In this exam-
tive with other cases. The final loads in Anchors 1, ple if considering only bending moment and wall
2 and 3 have the same tendency, i.e. quite un- movement, the “optimal” prestress load can be
changed until PR= 75%, afterward they increase chosen somewhere between PR= 100% and PR=
somewhat. The final load in Anchors 5, however, 125%. However with larger prestress load, the
increases almost linearly. The load in Anchor 4 has anchors and wailing beams need to be stronger and
an intermediate trend. more expensive. Considering this, the “optimal”
It can be expected that if a wall have a large PR should be chosen between 75% and 100%.
number of anchor levels, the final load of all the
upper anchors will have the same tendency as
Anchors 1, 2, and 3 in this example. This means
that no matter how the prestress load is chosen It is well known that bending moments obtained
from the beginning, at the final excavation stage from the traditional analysis is often overestimated
the anchor loads will reach to more or less the for the steel sheet pile walls. Performing a series of
same value. Choosing correct prestress load can model tests on sand of varying relative density,

Rowe (1952) showed that wall-soil interaction was Figure 10 shows the calculation of earth pres-
different for steel sheet piles and reinforced con- sure for the wall at Fredriksberg Station using the
crete wall due to the greater flexibility of the steel above method. In this study case σi = 34.5 kPa. If
sheet piles. This greater flexibility causes a redi- the maximum bending moment in the wall is com-
stribution of earth pressure, which differs consider- puted as for continuous spans, Mmax= (σi·hmax2)/10,
ably from the Rankine distribution. These changes where hmax is the largest distance between two ad-
reduce the design bending moment for a flexible jacent anchors, we will have Mmax= (34.5·3.52)/10=
pile wall. 42 kNm/m. If Mmax is computed as for simple span,
In Sweden the modified empirical strut load Mmax= (σi·hmax2)/8= (34.5·3.52)/8= 53 kNm/m. The
envelope method widely used, Ryner (1984), maximum bending moment obtained from Plaxis
the equivalent uniformly distributed pressure analysis in both cases with different prestress loads
σi = PA/(0.9∙H+d), where PA is total active earth is approximately 30 kNm/m, which is much lower
pressure up to the balance point, H is excavation than Mmax-values calculated according to the above
depth, and d is distance from the excavation bot- method.
tom to the balance point.

Figure 10. Fredriksberg Station - Calculation of earth pressure and wall bending moment according to traditional method

Figure 11. SL-10, South Link, Stockholm - Bending moment calculated according to different methods: 1) conti-
nuous spands using σi; 2) continuous spands using Rakine earth pressure; 3) FEM method by Plaxis

However, in a certain case, the empirical strut load tive” side, just behind the pre-stressed anchors,
envelope method can underestimate bending mo- soil behaves passively.
ments. One example is the case where sheet pile wall From the simple parameter study, it can be seen
is driven to bedrock. Let us examine the SL-10 that no matter how the prestress load is chosen
project, South Link in Stockholm. Figure 11 shows from the beginning, at the final excavation stage
the wall bending moment in the final stage for Case the load in anchors will reach to more or less the
100% obtained from Plaxis, and the maximum bend- same level. Choosing correctly prestress loads can
ing moment is 602 kNm/m. Using the empirical strut minimise considerably the wall and soil move-
load envelope method the maximum bending mo- ment, bending moment, as well as shear force in
ment can be computed as for simple span, Mmax= the wall.
(σi·Mmax2)/8 or for continuous spans Mmax= It is well known that bending moments obtained
(σi·Mmax2)/10. If take σi = 108kPa from Fig.11, we from the traditional analysis is often overestimated
have Mmax= 165kNm/m and 132 kNm/m respective- for the steel sheet pile walls. However, in a certain
ly. In Figure 11, the wall bending moments are also case, the traditional simplified method can unde-
calculated by different methods: 1) continuous beam restimate bending moment, such as the case of
under the equivalent earth pressure σi estimated ac- sheet pile wall driven to bedrock. In such cases, as
cording to the empirical strut load envelope method, a simplified calculation method, a frame analysis
2) continuous beam under Rankine active earth pres- with anchors modelled as inclined bars, may be
sure, and 3) Plaxis analysis, Case 100%. more realistic.
The maximum bending moment calculated by
the two simplified continuous beam methods are
120 kNm/m and 125 kNm/m respectively, in com- 6. REFERENCES
parison with the value obtained by Plaxis, 602
kNm/m. It is clearly that all the four Mmax-values ob- Fang, H-Y, et al (1991). Foundation Engineering
tained from the simplified traditional method are much Handbook, McGraw-Hill, N.Y.
lower than that obtained from Plaxis. This can be Phung Duc Long (2000). Soil stabilisation with
explained by the fact that in reality the supports of lime-cement columns - A solution for deep
the equivalent wall-beam are not fixed, while the sheet pile excavations in soft clay". Proc. Int.
continuos beam method assumes fixed supports at Conf. GeoEng 2000, Melbourne, Australian.
the anchor levels. We should also recall that the Phung Duc Long (2001a). Drilled-pile wall for
bending moment is a function of the beam dis- deep excavation at Frederiksberg Station, Co-
placement M(x)= E·y(2), and the fixities are primar- penhagen. Proc. 15th Int. Conf. on Soil Mechan-
ily important. For calculating a sheet pile wall dri- ics and Geotechnical Engineering, Istanbul,
ven to bedrock, other methods than the simple August, Vol. 2, pp 1147-1150.
span or continuous spans calculation should be Phung Duc Long (2001b). Some aspects in design
therefore used. In such cases, the frame method, of multi-anchored walls for deep excavations.
in which anchors are modelled as inclined bars, The 8th European PLAXIS users meeting, Bun-
should be more suitable. desanstalt für Wasserbau, Karlsruhe, Germany,
Phung Duc Long (2004). Why do we need FEM
5. CONCLUSIONS for design of multi-anchored walls for deep ex-
cavations? Proc. Vietnamese - Japanese Geo-
Full analysis of soil and wall stiffness and their technical Seminar VJSGE, 27-28 November,
interaction using realistic soil constitutive models Hanoi.
can be performed by means of FEM method. The Padfield C.J. and Mair R.J. (1984). Design of re-
principal advantages of such approach include the taining walls embedded in stiff clay. CIRIA
ability to model wall and soil deformation and Report 104, London.
stress in a realistic sequence of operation that fol- Rowe, P.W. (1952). Anchored sheet pile walls.
low actual construction stages. Prestress anchor Proc. Inst. Civ. Engrs, Part 1, Vol. 1, pp. 27-70.
load can also be taken into account in a realistic Ryner, A., Fredriksson, A. and Stille, H. (1996).
way. By the help of FEM method, some interest- Sheet pile wall handbook. Byggforsknings-
ing conclusions can be drawn bellow. rådet, Stockholm (in Swedish).
For the flexible sheet pile wall with anchor pre- Vermeer, P.A. et al. (1998). PLAXIS (Finite Ele-
stress load, the Rankine concept on active and pas- ment Code for Soil and Rock Plasticity) Version
sive side is no longer correct. At the so-called “ac- 7.1 - Users manual. A.A. Balkema, Rotterdam.